Composites Science and Technology 60 (2000) 1863±1872

www.elsevier.com/locate/compscitech

Flexural failure mechanisms in injection-moulded carbon

®bre/PEEK composites
$

Chi-Cherng Jeng, Ming Chen *

Institute of Materials Science and Engineering, National Sun Yat-Sen University, Kaohsiung, Taiwan 80424, ROC

Received 8 February 1999; received in revised form 15 September 1999; accepted 21 March 2000

Abstract
Two kinds of short-carbon-®bre-reinforced poly(ether ether ketone) composites were prepared with and without a transcrystal-
line interphase. The ®bre-length distribution, ®bre orientation and crystallinity were all characterized. These composites, with a
thickness of 3 mm and a span-to-thickness ratio of 32, were subjected to three-point-bending tests at cross-head speeds of 0.5, 5,
and 50 mm/min. Fractography was carried out by means of scanning electron microscope. Multiple shallow and deep shear cracks
were observed in the compression side of the damaged specimen. The appearance of the load/de¯ection curves and the fractography
suggest that compressive cracks occurred ®rst; the stress concentration under the loading nose and the shear stress maximum
induced the upper shear cracks which blunted the propagation of the compressive crack. The tension crack then initiated from the
tension side, met with the lower shear crack near the core region, and resulted in catastrophic failure. # 2000 Elsevier Science Ltd.
All rights reserved.
Keywords: A. Polymer-matrix composites (PMCs); A. Short-®bre composites; B. Mechanical properties; D. Scanning electron microscopy (SEM);
E. Injection moulding

1. Introduction transfer between the ®bre and the matrix. It is still an

open question, however, whether or not a transcrystalline
The most attractive features of short-®bre composites layer around the reinforcing ®bres in composites with
may lie in their potential for rapid, low-cost mass product- semi-crystalline thermoplastic matrices may improve the
ion and simplicity of manufacturing. Short-carbon-®bre mechanical property pro®le [1ÿ4].
poly(ether ether ketone) (SCF/PEEK) composites out- The mechanical properties of these composites are
perform samples/products made from the pure PEEK dependent on several important microstructural ele-
matrix in terms of sti€ness and strength. The carbon ments, including ®bre length, ®bre-volume fraction,
®bre improves the ¯exural de¯ection or the ¯exural ®bre orientation, ®bre/matrix interphase strength, and
crack resistance considerably. It is well established that morphological features such as degree of crystallinity,
the signi®cant improvement in mechanical properties spherulite size, lamella thickness and crystallite orientation
for reinforcement in semi-crystalline polymers over [5,6]. These features are, in turn, a€ected by variations in
amorphous polymers can be partly attributed to the the processing conditions, such as mould geometry, melt
morphology and crystallinity of the polymer matrix in the temperature, moulding temperature, and the rheological
interfacial region. Fibres a€ect the matrix morphology in properties of the moulding compound. Information
semi-crystalline melts because the ®bres can act as about the e€ect of the transcrystalline interphase on the
nucleation rods causing a structure known as trans- macroscopic performance of long-®bre-reinforced
crystalline. It is believed that the transcrystalline interphase, polymers with good ®bre/matrix adhesion is available.
a kind of `physical coupling,' would enhance stress Spherulitic and transcrystalline morphologies of injection-
moulded ®bre-reinforced samples are dicult to inves-
$
Presented at 12th International Conference on Composite Mat- tigate because of preparation problems. The increased
erials (ICCM-12), in Paris, France, 5±9 July 1999.
* Corresponding author. Tel.: +886-7-525-2000-4062; fax: +886- use of thermoplastics has brought about a need for better
7-525-4099. understanding of the processing techniques used to
E-mail address: mingchen@mail.nsysu.edu.tw (Ming Chen). manufacture these materials.
0266-3538/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0266-3538(00)00076-2
1864 C.-C. Jeng, M. Chen / Composites Science and Technology 60 (2000) 1863±1872

In previous papers [7,10], thermal stability experi- radius of the cylindrical surfaces of the nose and sup-
ments [7] have indicated that unreinforced PEEK was ports was 3.2 mm. The plaques were tested at room
stable in nitrogen up to a temperature of 400 C, holding temperature (RT) on an Instron universal testing
for 15 min. This treatment was e€ective in reducing machine (Model 1125) at crosshead speeds of 0.5, 5 and
nucleation density to the point of allowing morphology 50 mm/min, respectively. The de¯ection reading as a
control [8ÿ10]. In this paper, attention is focused on the function of the applied load was recorded during the
e€ect of ®bre orientation and transcrystalline interphase entire test until the applied load dropped substantially,
on the ¯exural properties of the composites. Attempts as a result of internal damage in the specimens. The
have been made to study the ¯exural failure mechanism fracture surfaces of broken test bars were cold sputtered
of SCF/PEEK composites. with a very thin layer of gold and were examined with a
JEOL JSM 6400 scanning electron microscope (SEM)
with a high voltage of 20 kV for the secondary electron
2. Experimental image.

Injection-moulding grade composite pellets were

ordered from RTP company under the trade name RTP 3. Results and discussion
2285. In this material, PEEK of grade 150 is forti®ed
with 30 wt.% SCF. The pellets were melted in a heated 3.1. Microstructure characterization
cylinder at 400 C and were injected at a mould tem-
perature of 180 or 120 C using a machine of standard Tadmor [12] treated the front region of mould ®lling,
con®guration (Engel ES 200/50). The samples have the and proposed a model to explain the experimental
following codes: 400/180 and 400/120. The number observations. Both the high degree of orientation at the
before the slash (400) indicates the melted temperature surface and no orientation at the centre are originated
at the cylinder and the number after the slash (180 or from the fountain type ¯ow [13] in the advancing front
120) stands for the mould temperature. The ¯exural region. The moulding-induced microstructure of the
bars of 127 mm12.5 mm3 mm (according to the short ®bre reinforced grades can be approached by the
ASTM D790 standard [11]) were injection moulded and above ¯ow model of Tadmor [12] and Rose [13]. The
characterized in terms of the ®bre orientation, core size, elongational and shear ¯ow ®elds developed during the
average ®bre length, ®bre-length distribution and degree cavity ®lling process are responsible for the well-known
of crystallinity. three-layer ®bre orientation microstructure. The ®bres
Specimens of 10 mm12.5 mm3 mm were cut from were highly aligned to the MFD (mould-®lling direction)
the centre part of the bar. The transverse sections and in the two surface layers, whereas ®bres in the central
the surfaces of the specimens were embedded in a phe- layer were dispersed perpendicular to the MFD. The
nolic resin and polished successively down to a 0.05 mm area of the central layer was 0.8% of the cross-section in
aluminum oxide ®nish. Results on the ®bre orientation this study, i.e. the contribution of the core region was
and the core size across the bar thickness were estab- rather low in these 3 mm thick PEEK composites. The
lished by re¯ected light microscopy (LM) of the polished surface from the 400/120 specimens along the
polished surfaces. The ®bre-length distribution curves of MFD showed a high value of ®bre orientation (Fig. 1a),
®ve 6ÿ10 mg pieces, cut along the thickness direction, which was consistent with the characteristic feature of
were determined by transmitted LM after burning o€ the maximum orientation at the surface. The velocity
the PEEK matrix at about 500 C under a dry air ¯ux in pro®le, at any ®xed location, was maintained until the
a thermogravimetric analyzer (Perkin-Elmer TGA-7). abrupt cessation of ¯ow when the mould was ®lled at a
Randomly selected cross sectional views were photo- mould temperature of 120 C. However, the short ®bres
graphed at an e€ective 100 magni®cation. The average showed an in-plane random distribution on the surface
®bre length of at least 350 ®bres obtained from four to (Fig. 1b) of the 400/180 specimens. This exterior layer
®ve images was measured statistically for each speci- was about 200250 mm thickness through the moulding
men. The overall crystallinity of the specimens across thickness or the transverse section from the observation
the bar thickness was detected by di€erential scanning of re¯ected LM. Upon the cessation of ¯ow, relaxation
calorimetry (DuPont 910 DSC) at 20 C/min heating of the melt [14] started and was followed by a dis-
rate. Both temperature and heat-¯ow scales were cali- orientation of the ®bres because the mould temperature
brated with indium and lead at the same heating rate at 180 C was beyond the Tg of PEEK (143 C). Therefore,
under a constant nitrogen ¯ux. the 400/180 specimens displayed ®ve-layer structure
The mechanical test used was the three-point-bending across the thickness.
test with a span-to-thickness ratio (s/t) of 32:1 (ASTM Fig. 2a and b show the ®bre-length distribution of
D790 [11], Test Method I). in order to avoid excessive 400/120 and 400/180 specimens, respectively. Ln is the
indentation, or failure under the loading nose, the number average of ®bre length, calculated by
 C.-C. Jeng, M. Chen / Composites Science and Technology 60 (2000) 1863±1872 1865

Fig. 1. Re¯ected light micrographs from the polished sections of the surface of (a) 400/120, (b) 400/180 specimens. The scale bar corresponds to
100 mm.

Fig. 2. The ®bre-length distribution of (a) 400/120, (b) 400/180 specimens.

P
Ni Li layer was completed in 30 s at 180 C during the holding
Ln ˆ P
Ni and cooling periods of injection. The apparent fusion
enthalpy was estimated by subtracting the exothermic
Lw is the weight average of ®bre length, calculated by recrystallization peak area from the melting peak area.
P This measure was quite inexact because the fusion
Ni L2i enthalpy and the heat of crystallization per unit crys-
Lw ˆ P
Ni L i tallinity were quite di€erent near 170 C from that at
330 C [9,10]. The 70% of crystallinity of the composite
where Ni is the numbers of ®bres having length Li. No is determined as the ratio of the apparent fusion
discernible di€erence of the average ®bre length across enthalpy to the fusion enthalpy of a sample of PEEK
the bar thickness was found among the specimens. The with 100% crystallinity, taken as 130 J/g [15]. The crys-
Ln for specimens moulded at 120 and 180 C were 118 tallinity contents of 400/180 and 400/120 specimens are
and 107 mm, respectively. The corresponding Lw were thus 28 and 24%, respectively. The latter was probably
147 and 145 mm. The ®bre length distributions (Lw/Ln) over estimated because of recrystallization [9,10].
were 1.25 and 1.36, respectively.
All the DSC traces from various depths of the thick- 3.2. Flexural properties at room temperature
ness of the 400/120 test bars showed the recrystalliza-
tion behaviour. The cooling rate of the central layer was Fig. 3a and b shows the load/de¯ection curves for the
also high at a mould temperature of 120 C. There was 400/120 and 400/180 specimens, respectively, tested at
no recrystallization at all in the case of 400/180 speci- RT and at a cross-head speed of 0.5 mm/min. The load
mens. This indicates that the crystallization of the skin increased up to an average de¯ection of 9.0 mm for the
1866 C.-C. Jeng, M. Chen / Composites Science and Technology 60 (2000) 1863±1872

Fig. 3. Typical load/de¯ection curves for (a) 400/120, (b) 400/180 specimens tested at room temperature and at a cross-head speed of 0.5 mm/min.

400/120 specimens, and 11.4 mm for the 400/180 speci- be ascribed to the higher ®bre orientation of 400/120
mens. These de¯ections corresponded to the maximum specimen. Also, the modulus of the composites increases
stress level without any initial load drop and the load with crosshead speed, which is characteristic of the
dropped o€ sharply as soon as the fracture was initiated common strain-rate dependence of polymers. The mod-
on the tension side. This implies that the composite was ulus values obtained at RT (<Tg) give an indication of
very brittle without any crack-propagation step at RT. the global degree of ®bre orientation, but do not re¯ect
Crack initiation was thus believed to be the dominant the degree of crystallinity.
failure energy absorption process of fracture for T<Tg. The corresponding de¯ection and strength are also
The composite responses between the applied load and listed in Table 1. For all cases shown, the de¯ection of
the de¯ection were quite linear in the ®rst 3.5 mm, after the 400/180 specimens was about 20% higher than that
which the curves displayed nonlinearity. In the non- of the 400/120 specimens. The de¯ection values
linear region, failure mechanisms on a microscale (e.g. obtained at RT (<Tg) are seen not to be sensitive to
debonding at the SCF/PEEK interface) were possible. percent crystallinity. Unlike the de¯ection, the strength
From the load/displacement curves, the ¯exural of the 400/180 specimens was always slightly higher
modulus and the ¯exural strength were calculated by than that of the 400/120 specimens. The strength values
using the equations of homogeneous beam theory. in MPa increased sharply from a speed of 5 to 50 mm/
Table 1 lists the ¯exural modulus as a function of min for both 400/120 and 400/180 specimens. The e€ect
crosshead speed. The modulus in GPa for the 400/120 of cross-head speed was in the opposite direction to that
and 400/180 specimens, respectively, were found to be observed for other neat polymer systems, i.e. a trend to
24.2 and 21.9 at 0.5 mm/min; 24.2 and 22.3 at 5 mm/ brittle fracture with increasing strain rate. The major
min; 24.9 and 22.8 at 50 mm/min. Each value represents di€erence is considered to be a result of crystallization
an average of at least three specimens. A rise in ¯exural on the carbon ®bres, which possibly contain tran-
modulus in the 400/180 specimen containing higher scrystalline regions. At RT, ¯exural behaviour is in¯u-
crystallinity is expected. However, the data illustrated in enced by the microstructural parameters of ®bre
Table 1 do not con®rm this expectation. This can probably orientation and transcrystallinity in composites.

3.3. Failure path and mechanisms

Table 1
Flexural properties of SCF/PEEK composites versus cross-head speeda
3.3.1. Fracture behaviour
Specimen code 400/120 400/180 As shown in Fig. 3, the load dropped sharply at the
Speed (mm/min) 0.5 5 50 0.5 5 50 maximum load, which corresponds to the tensile failure
in the outer surface. This implies that the composite was
Modulus (GPa) 24.2 24.2 24.9 21.9 22.3 22.8
very brittle and without any propagation step at RT.
(0.4)b (0.9) (0.8) (0.3) (0.4) (0.2)
The SEM micrographs at longitudinal section through
De¯ection (mm) 9.0 9.0 9.5 11.4 10.9 11.3 the centre of the damaged 400/120 and 400/180 speci-
(0.5) (0.4) (0.3) (0.4) (0.2) (0.7)
mens, tested at a crosshead speed of 5 mm/min, are
Strength 334.1 338.6 359.0 356.2 357.5 375.7 presented in Fig. 4a and b, respectively. For both spe-
(9.9) (9.4) (7.3) (2.1) (3.9) (15.8)
cimens, damage in the form of shallow crack appeared
a
Three-point-bending test (s/t=32) at room temperature. on the compression side just (about 250 mm) below the
b
The number inside the parentheses is the standard deviation. loading nose (arrow A in Fig. 4a and b), this shallow
 C.-C. Jeng, M. Chen / Composites Science and Technology 60 (2000) 1863±1872 1867

Fig. 4. SEM micrographs of the fracture surface of (a) 400/120, (b)

400/180 specimens tested at a cross-head speed of 5 mm/min. The scale
bar corresponds to 1 mm. Top and bottom edges are presented as
black.
Fig. 5. SEM fracture micrographs of the convex part of the 400/180
specimen: (a) top side (b) bottom side. The specimen was tested at a
crack did not take place over the entire specimen width. cross-head speed of 0.5 mm/min. The scale bar corresponds to 10 mm.
One much deeper crack was recognized (arrow B in Fig.
4a and b) between the core and the shallow crack.
SEM micrographs of a close view of both sides of the the compressive crack occurred ®rst (see Section 3.3.4).
deeper crack at a tilt angle of 23 (one side was rotated It is also suggested that the deep shear crack on the
180 ) are presented in Figs. 5a and b and 6a and b for a compression side occurred before the tensile crack initiated
damaged 400/180 specimen tested at a cross-head speed from the tension side. Based on the above observation, the
of 0.5 mm/min. The schematic location of Figs. 5 and 6 shear cracks along the longitudinal direction through
is given in Fig. 7. The direction of plastic drawing the centre of the damaged specimen are sketched in Fig.
(arrow direction) is opposite between the pair of the 7. This ¯exural failure behaviour is quite di€erent from
convex and concave parts, and the pair of the top and that of continuously unidirectional composites, where
bottom sides of the crack. Shear cracks on the top and the tensile failure dominates at a span ratio of 32:1.
bottom sides were recognized from the opposite draw-
ing direction of the matrix. Thus, the crack propagation 3.3.2. Transcrystalline e€ects
on the top side is independent from the crack formed on SEM micrographs in Figs. 8 and 9 show the fractured
the bottom side. The fracture surfaces show ®bres sur- surfaces of PEEK reinforced by SCF at a cross-head
rounded by matrix material in both sides of the fracto- speed of 0.5 mm/min. In particular, Fig. 8 (500 mag-
graphy. This indicates that the composite did not fail ni®cation) refer to sample 400/120 and Fig. 9 (3000) to
adhesively, but cohesively. sample 400/180. As the load increased, elastic strain
Signi®cant damage including multiple shallow and energy was accumulated in the specimen, then at large
deep shear cracks was found right after the load drop, de¯ection, a large number of ®bres were pulled out quite
indicating that the damage occurrence and growth were cleanly on the tension side (Figs. 8c and 9c). The crack
progressive in the initiation step. The 7.5 mm non-linear propagated along the ®bre/matrix interface and ®bre
region indicates that stable crack propagation constitutes ends, leaving bare ®bres on the failure surface. No
the major part of the initiation step. The appearance of the strong di€erences in SCF/PEEK adhesion were evident
load/de¯ection curve and the fractography suggest that on the tension side.
1868 C.-C. Jeng, M. Chen / Composites Science and Technology 60 (2000) 1863±1872

Fig. 7. Schematic of deep shear crack across the longitudinally

damaged specimen.

In Figs. 8b and 9b, SEM micrographs of the core

Fig. 6. SEM fracture micrographs of the concave part of the 400/180 region show that ®bres were embedded in the PEEK
specimen: (a) top side (b) bottom side. The specimen was tested at a
cross-head speed of 0.5 mm/min. The scale bar corresponds to 10 mm. matrix Ð cohesive matrix failure. It clearly indicates
both good adhesion and insucient matrix shear
strength. The transcrystalline regions were found across
almost all the ®bres observed under higher magni®ca-
Bare ®bres were also observed on the compression tion (3000) in the cases of the 400/180 specimens tes-
side for the specimens moulded at 120 C (Fig. 8a). No ted at crosshead speeds of 0.5, 5 and 50 mm/min. For
marked di€erence in the extent of pull-out of the ®bres the 400/120 specimens, in contrast, an increase of speed
was observed between the compression and tension from 0.5 to 50 mm/min resulted in a change from cohe-
sides. If the homogeneous nucleation is dominant and sive failure to adhesive failure. The results showed clearly
incomplete, impurities will be accumulated at the ®bre- that the occurrence of transcrystallization was depen-
matrix interface and the interfacial bond will be weak dent on the specimen cooling rate which was slower at
[5,16]. This implies a poor bonding between SCF and higher moulding temperature (180 C>Tg>120 C) and
PEEK matrix with a fully spherulitic morphology in the in the core region. In the section of microstructure
400/120 specimens. In contrast, the fracture surface for characterization, the 400/120 specimens showed the
the specimens moulded at 180 C showed ®bres sur- recrystallization behaviour in the core region which
rounded by a sheath undeformed matrix material (1 m) indicated that the crystallization of the 400/120 specimen
on the compression side (Fig. 9a). According to the was incomplete during the holding and cooling periods
20% higher de¯ection and higher strength at break of injection. The matrix shear strength was reduced at
values (see Table 1), this is indicative of a strong bond- higher strain rate because the amount of amorphous
ing between the ®bres and matrix which can be attrib- PEEK was high. In the 400/180 specimens, the tran-
uted to the presence of the transcrystalline interphase. scrystalline region resulted in the 20% higher de¯ection
The boundaries between spherulites and transcrystalline and the corresponding enhanced strength.
regions provided the weak fracture paths in the material
because of the di€usion of low molecular weight and/or 3.3.3. Failure path
noncrystallizable fractions to the growing crystallization There are several possibilities for the occurrence of
fronts [5,16]. the transcrystallization: (1) slow cooling after melting at
 C.-C. Jeng, M. Chen / Composites Science and Technology 60 (2000) 1863±1872 1869

high temperature to remove the nuclei [8,9]; (2) stress much slower than that in the surface layer because of
induced crystallization [17,18]; (3) the stresses induced the low thermal conductivity of PEEK matrix. Fig. 10
by the di€erence in thermal expansion coecients shows a spherulitic pattern (arrow) [21] observed in the
[19,20]. In the surface layer, the shear stress was max- core of the 400/120 specimen after the ¯exural test at the
imum and the mismatch in thermal expansion coe- speed of 5 mm/min. The spherulitic size within the
cients was also maximum during cooling to the mould matrix was estimated to be 5 mm in diameter. It took
temperature. In the core region, the cooling rate was 24 and 13 s, respectively, to grow this spherulite if the

Fig. 8. SEM micrographs of the fracture surface of 400/120 specimen: Fig. 9. SEM micrographs of the fracture surface of 400/180 specimen:
(a) compression side, (b) core, and (c) tension side. The cross-head (a) compression side, (b) core, and (c) tension side. The cross-head
speed was 0.5 mm/min. The scale bar corresponds to 10 mm. speed was 0.5 mm/min. The scale bar corresponds to 10 mm.
1870 C.-C. Jeng, M. Chen / Composites Science and Technology 60 (2000) 1863±1872

corresponding crystallization temperatures were 310 unrealistic to refer to the transcrystallized interphase
and 305 C [8]. It implies that the nuclei surviving in the and the spherulite matrix as the interphases now
core were much less than expected and the cooling rate impinge on each other to form the major fraction of the
was slower, too. Therefore, it was possible to allow the matrix in the 400/180 specimens. These boundaries
growth of the transcrystalline layer in the core of the provided weak fracture paths in the material. In this
400/120 specimens. Fig. 11 shows a hackle-like pattern study, direct evidence for better interfacial bonding of
(arrow) between two ®bres in the core of the 400/180 transcrystallinity to the ®bres can be illustrated by
specimen where the direction of crack propagation was microscopy observation, as shown in Fig. 9a and b.
perpendicular to the applied force. This pattern also Transcrystalline structures combined with good ®bre/
supports the shear force applied in the core region. matrix adhesion changed the fracture path. A large
Three di€erent fracture paths are possible [16]: (1) at amount of transcrystalline material meant also an easier
the SCF/PEEK interface; (2) at the interphase between transfer of stress, and thus the materials are able to
spherulitic and transcrystalline structures; (3) through sustain large de¯ections before the break. Based on the
the PEEK matrix by cohesive failure. The arrangement above observation, the ¯exural failure path (s/t=32) is
of boundaries between spherulites and transcrystalline schematically presented in Fig. 12 for the 400/180 spe-
regions is dependent on the relative distribution of the cimens tested at RT and at a cross-head speed of 0.5
two phases and on the strength of the interfaces between mm/min.
the two phases. It should be noted that it is now somewhat
3.3.4. Failure sequence
Four types of damage developed in the previous
results. They are compression, upper shear, lower shear,
and tension cracks. The sequence of damage progres-
sion is emphasized in this section. If the deep shear
crack occurred ®rst, the damaged portions of the speci-
men may continue to act as separate beams, and the
energy absorbed in this section is rather low. Thus, the
deep shear crack is not supposed to precede both com-
pression and tension cracks.
Now compare the sequence of compression and ten-
sion cracks. For the span to thickness ratio (s/t=32)

Fig. 10. Spherulitic pattern observed in the core of 400/120 specimen.

The cross-head speed was 5 mm/min. The scale bar corresponds to 10
mm.

Fig. 11. Hackle-like fracture observed in the core of 400/180 speci- Fig. 12. Schematic ¯exural failure path for 400/180 specimen with
men. The cross-head speed was 50 mm/min. The scale bar corresponds transcrystalline layer. Three-point bending test (s/t=32) at room tem-
to 10 mm. perature and at a cross-head speed of 0.5 mm/min.
 C.-C. Jeng, M. Chen / Composites Science and Technology 60 (2000) 1863±1872 1871

used in these experiments, the ¯exural failure mode was the upper shear crack. In addition, the upper shear
expected to be of the tensile type (instead of shear fail- crack did not take place over the entire specimen width.
ure). If the tension crack preceded the compression The s/t ratio, stresses and de¯ection of the composite
crack, the deep shear crack would be on the tension side would then be recalculated at the same load, and the
and would fail in Mode I where the direction of plastic propagation of the upper shear crack is restricted, too.
drawing was parallel in both sides of the crack. This is The upper shear crack may also be stopped by the dis-
in disagreement with the observations. From the orientation of the ®bres near the core region [24].
appearance of load/de¯ection curves and the fracto- There are two possibilities regarding the relation
graphy, considerable damage appeared on the compres- between the lower shear crack in the middle layer (#3)
sion side just below the loading nose, thus the and the tension crack in the bottom layer (#4). Either
compressive crack is suggested to occur ®rst (denoted as the lower shear crack appeared ®rst, and stopped the
#1 in Fig. 13). upper shear crack from spreading sideways, or the
According to numerical analysis, the ®rst compression upper shear crack itself diverted into the tension crack.
crack in the middle layer should start right under the In all of these tests, a tensile failure of the outermost
loading nose because of the stress concentration [22]. It ®bres on the tension side was observed at the point
is possible that the initiation of the upper shear crack where the maximum load was reached. During the
(denoted as #2 in Fig. 13) was a consequence of this loading up to catastrophic fracture, the shear micro-
initial compressive failure or of the maximum shear stress cracks induced matrix macrocracking and caused pull-
[23]. The compression cracks propagated downward, and out of the ®bres from matrix. Thus the lower shear
interacted with the expanding upper shear cracks because crack and the tension crack can be regarded as mutually
the shear stress maximum also moved towards the core exclusive events happening almost together. Direct evi-
region. The inclination of the deep cracks also suggests dence in producing these cracks is required.
the possibility of shearing action producing the cracks. Based on the above discussion, the events (compres-
The above discussion precludes the possibility of upper sion crack, upper shear crack, lower shear crack, and
shear cracks originating from the lower tension cracks. tension crack) leading to failure of the injection-moul-
Thus, compression crack and the upper shear crack pre- ded SCF/PEEK composites are schematically presented
ceded the tension crack in the bottom layer. The com- in Fig. 13.
pression crack, associated with the predeformation of the
upper shear crack, caused the ductile failure with a cor-
responding large de¯ection of about 9±11 mm for injec- 4. Conclusions
tion-moulded SCF/PEEK composites.
The upper shear crack's propagation and further In this study, a change of 60 C in the mould tem-
compression cracking were related to each other. The perature (one above Tg and one below Tg) resulted in a
compression crack away from contact zone was almost change from a transcrystalline morphology to a
vertical. With further loading, the compression crack spherulitic morphology. The transcrystalline interphase
(#1) and the upper shear crack (#2) described above enhanced the ®bre/matrix adhesion, improved the ¯ex-
joined together to initiate a macroscopic fracture. The ural de¯ection and strength of the composite. At room
propagation of the compression crack was blunted by temperature, the ®bre orientation dominated the mod-
ulus. Multiple shallow and deep shear cracks were
observed in the compression side of the damaged speci-
mens. The compressive crack is suggested to occur ®rst;
the stress concentration under the loading nose and the
shear stress maximum induced the upper shear crack
which blunted the propagation of compressive crack.
Then the tension crack initiated from the tension side,
met with the lower shear crack near the core region, and
resulted in catastrophic failure. The transcrystalline
interphase changed the ¯exural failure path and was
able to sustain larger de¯ections before the break.

Acknowledgements

Fig. 13. A schematic drawing of the failure sequence of the injection-

This work was supported ®nancially by the National
moulded SCF/PEEK composites. Three-point-bending test (s/t=32) Science Council, ROC, under contract numbers NSC
at room temperature and at a cross-head speed of 0.5 mm/min. 85-2216-E110-003 and NSC 87-2216-E110-011.
1872 C.-C. Jeng, M. Chen / Composites Science and Technology 60 (2000) 1863±1872

References and reinforced plastics and electrical insulating materials (ASTM

D790-90), vol. 08.01. Philadelphia (PA): American Society For
[1] Sarasua JR, Remiro PM, Pouyet J. The mechanical behavior of Testing and Materials, 1992. p. 272±281.
PEEK short ®bre composites. J Mater Sci 1995;30:3501±8. [12] Tadmor Z. Molecular orientation in injection molding. J Appl
[2] Karger-Kocsis J, Friedrich K. Microstructure and fracture Polym Sci 1974;18:1753±72.
toughness of short ®bre reinforced injection-moulded PEEK [13] Rose W. Fluid-¯uid interfaces in steady motion. Nature
composites. Plast Rubb Process Appl 1987;8:91±104. 1961;191:242±3.
[3] Xu J, Zhang M, Zhang Z, Xiong X, Zeng H. The tensile behavior [14] Lopez LC, Cieslinski RC, Putzig CL, Wesson RD. Morphologi-
and failure of short carbon ®ber reinforced PEEK composite with cal characterization of injection moulded syndiotactic poly-
transcrystalline interphase. In: Proceedings of the tenth interna- styrene. Polymer 1995;36:2331±41.
tional conference on composite materials, vol. 4, 1995, Whistler, [15] Blundell DJ, Osborn BN. The morphology of poly(aryl-ether-
BC, Canada. Cambridge (UK): Woodhead Publishing, 1995. p. ether-ketone). Polymer 1983;24:953±8.
101±8. [16] Hull D. Matrix-dominated properties of polymer matrix compo-
[4] Tregub A, Harel H, Marom G. The in¯uence of thermal history site materials. Mater Sci Eng A184:173±83
on the mechanical properties of poly(ether ether ketone) matrix [17] Kobayashi K, Nagasawa T. Crystallization of sheared polymer
composite materials. Compos Sci Technol 1993;48:185±90. melts. J Macromol Sci-Phys B4:331±45
[5] Lee Y, Porter RS. Crystallization of poly(etheretherketone) [18] Yeh GSY, Hong KZ. Strain-induced crystallization: part III.
(PEEK) in carbon ®ber composites. Polym Eng Sci 1986;26:633±9. Theory Polym Eng Sci 1979;19:395±400.
[6] Avella M, Martuscelli E, Pascucci B, Raimo M. Tran- [19] Campbell D, Qayyum MM. Melt crystallization of poly-
scrystallinity phenomena in ®ber-reinforced polypropylene: II. propylene: e€ect of contact with ®ber substrates. J Polym Sci
Morphology, thermal and mechanical properties relationships Polym Phys Edn 1980;18:83±93.
Polym Eng Sci 1992;32:383±91. [20] Thomason JL, Van Rooyen AA. Transcrystallized interphase in
[7] Chen JY, Chen M, Chao SC. Thermal stability and crystal- thermoplastic composites: part II. In¯uence of interfacial stress,
lization kinetics of poly(ether ether ketone). Macro Chem Phys cooling rate, ®bre properties and polymer molecular weight. J
1998;199:1623±9. Mater Sci 1992;27:897±907.
[8] Chen M, Chen JY. Analysis of crystallization kinetics of poly(ether [21] Purslow D. Matrix fractography of ®bre-reinforced thermo-
ether ketone). J Polym Sci Part B Polym Phys 1998;36:1335±48. plastics: part 1. Peel failures Composites 1987;18:365±74.
[9] Chen M, Chao SC. Thermal stability and nonisothermal crystal- [22] Joshi SP, Sun CT. Impact induced fracture in a laminated com-
lization of short ®ber reinforced poly(ether ether ketone) compo- posite. J Compos Mater 1985;19:51±66.
sites. J Polym Sci Part B Polym Phys 1998;36:2225±35. [23] Sullivan JL, Van Oene H. An elasticity analysis for the generally
[10] Chao SC, Chen M, Chung CT. Isothermal crystallization and and specially orthotropic beams subjected to concentrated loads.
melting behavior of short ®ber reinforced poly(ether ether Compos Sci Technol 1986;27:133±55.
ketone) composites. J Polym Res 1998;5:221±6. [24] Liu D. Impact-induced delamination Ð a view of bending sti€-
[11] Anon. Standard test methods for ¯exural properties of unreinforced ness mismatching. J Compos Mater 1988;22:674±92.